Jim Keough

@sama If you are saying same intelligence and different price/speed, then yes

i get some anxiety not using the smartest-available model/settings. but sometimes i dont mind if it's really slow. i wonder if we should focus more on a price/speed tradeoff relative to a price/intelligence tradeoff.

@philz1337x @replicate very expensive on replicate

Now live on @replicate 🤩 replicate.com/philz1337x/cla…

✨ Clarity Upscaler now has a Pro version It's the first upscaler that can add realism without changing a person's identity Two years ago I launched Clarity Upscaler, which has since made 30 million runs. I still can't really believe it 🤩 It's great for creative upscaling but it always changes people's identity too much. I made Crystal to solve that, but it's not able to add details or realism. Now we have Clarity Pro Upscaler which finally solves all that and can enhance images without changing the content. In the coming days I will make it even more powerful by adding parameters and increasing the max resolution even further (it's already at 8K!) Happy to hear your feedback

@BoyuanChen0 Gpt-image-2 is not able to preserve small/blurry text. In this example, it changes a house number. I tried several prompts. Overall, it is incredible though!

We are committed to continually improving the GPT Image 2 model! I am actively fixing various issues from the community feedback. Just reply or DM me your GPT conversation! Features like 2K or 4K images are already available via the experimental API. Hope you enjoy the model!

@herrmanndigital Hearing this from so many brands. Cc @JoeJesuele

This week, like many other weeks in 2026 (2025 / 2024 / 2023 / 2022...) Meta ads simply broke. You probably haven't heard a peep from Meta on it, but it's been brutal. We're talking turning ad accounts off brutal. Some of you were probably 100% immune. Others had no idea because your teams suddenly tried shifting their creative / offers. Others work with agencies that probably set it and forgot it. Whatever it may be... I sound like a broken record here venting. But this just simply is becoming more and more unacceptable at the sheer level of these bugs. There feels like no recourse. No post-mortems. No true apologies. Just simply, break then kill a brand's performance for around a week, get a small refund, then move on to the next time it breaks. I'm tired guys. I know X is here to post about AI automations and how awesome I am, but the reality is this week (much like many other weeks prior) I find myself being therapist to CMOs and brand teams trying to figure out is this a bug, a creative issue, or the business. More and more it's a bug. I want Meta to be stable, but every year from Feb - May it's just a mess. Do we simply just pull budget during Meta's annual bug season? I'm out of answers personally but on behalf of the brands that pay me to manage advertising across every ad platform, I'd like some answers from Meta because it's the only one plagued with these issues at this scale.

@AnthonyHigman @PPCGreg @adsliaison Concur

@PPCGreg @adsliaison Actually yes this is true. Different levels of targeting work differently. Its in the help docs. We always go top down and layer in, but this is actually accurate.

Had an xwf.google rep tell us targeting by City or County "helps the ads trigger more effectively and reach more people." vs targeting by zip code - does adding "Florence, SC" as a target city here help if we already have all of the zipcodes tagged? @adsliaison

@JoeJesuele @HomeJab Putting the Agent 🤖 in Real Estate Agent

Last week was interesting. Marc Benioff announced Salesforce is going headless. APIs as the UI, agents as the users. Two days earlier, we released @HomeJab's API and MCP. Same shift. But here’s the difference: Salesforce → AI operates software HomeJab → AI executes in the real world Agents can now order photography, schedule jobs, and receive media, all without a single form. The next phase of AI isn’t just managing software. It’s doing real work.

@jasonyimco Does this help if running meta ads instant forms (capi for CRM events) ?

Meta just killed the biggest technical barrier in performance marketing. for years, the gap between big brands and small businesses wasn’t just budget. it was developer resources. if you wanted to run dynamic product ads or set up the conversions api, you needed a dev team to configure servers and manually map structured data. that changes today. we just rolled out two massive updates that automate the entire backend setup: 🔵 ai-powered pixel enrichment: the pixel now uses ai to automatically read your website and pull product names, prices, and availability into your events. no more manual coding or schema .org mapping. smaller businesses get performance benefits without technical work and larger businesses can refocus tech resources on more important areas 🔵 one-click capi: we launched a “meta-enabled” conversions api setup. it’s literally one click. no servers, no ongoing maintenance, no costs. this is a huge deal for signal quality. advertisers using capi for web events see an average 17.8% lower cpa than those relying on the pixel alone. historically, getting that setup was a nightmare for lean teams. now, the playing field is leveled. the machine handles the technical plumbing, so you can focus entirely on creative and strategy. if you’ve been putting off capi because it was too technical, now is the time.

@samuel_spitz Can Replit make images in the 3 main aspect ratios and export in that ratio?

@KodyNordquist This is good. I would use it.

Meta is rolling out per ad spend allocations. Very slow roll out - do you think this will be useful or is it providing people more than they need/will use?

@samuel_spitz @JamesonCamp Oh yeah I noticed that today. What's the difference? Are those all skills that Replit built directly?

@JimLeadGen @JamesonCamp Press the + button from the input box, rather than going to this pane

We are 99% of the way to letting generalists and ecom people execute $50k of dev work and $10k of design work in a weekend AI closed the coding gap a while ago. The problem was always design. Your code worked...it just looked like a developer built it. I've been using Manus to design Shopify funnels and it's actually great for that. But the second you try to edit a specific element, you're basically starting over. Replit just cracked this: 1. Tell it what you want 2. It spins up 3 design variations at once 3. Pick the one you like most 4. Actually edit the individual elements without the whole thing breaking Step 4 is the unlock. Nobody cared that AI could generate a landing page. We cared that we couldn't touch it after without blowing it up. Oh and by the way, you can have multiple agents building in parallel while you're editing. So while you're dialing in the design on one page, the next three are already being built. If you're a Figma power user this won't move you. But if you're a small team building funnels, landing pages, or apps and you know the loop of generate, hate the design, regenerate, lose your edits... That loop is over.

@samuel_spitz @JamesonCamp Where is the ad creative skill?

@JamesonCamp Try the new ad creative skill!

@samuel_spitz @JamesonCamp Is there a Replit Animation Skill?

@amasad The complexity looks insane. I'm worried about the forking and multiplayer being stable 🤞 a little nervous to actually use it.

Any interest in the tech details behind agent 4? We had to solve gnarly problems around forking and merging entire compute environments to run parallel agents. We also had to sort our dependencies, do inter-task and cross-task parallelizing, and make all of that multiplayer.

@JoeJesuele Let's go!

AI video for real estate is improving incredibly fast. We're just getting started.

@codyplof Coupler.io I think is best if you aren't getting super complex. It's plenty for most needs and much cheaper. Other options: airbyte, supermetrics, fivetran.

What is everyone using for their data stack? We’re considering making a move and most in need of a better ETL solution

@samuel_spitz Yes please

Some big stuff coming soon Let me know if you want an early look

@okuiux @supercut_video Wow, this is awesome

Just shipped 3 big updates to the @supercut_video AI assistant. 👁️ Vision —it can now see your videos for way better context 🖼️ Frame pulling —surfaces relevant frames to illustrate its answers 🤖 New agent skill —record your screen & auto-generate a skill.md

You don't need a $2K/month attribution platform. Just ask AI to write 2 Google Tag Manager tags. Tag 1 fires on all pages. It reads your URL parameters — UTMs, gclid, fbclid, whatever you're tracking — plus the referrer, and stores everything in first-party cookies. First touch gets locked in on the initial visit. Last touch gets updated every time someone comes back through a new campaign. Cookie sits on your root domain so it works across subdomains. Tag 2 fires on your conversion page. Thank you page, order confirmation, whatever yours is. It reads those cookies, grabs the order ID and revenue from your dataLayer, and sends the whole thing to a webhook. What you do with the webhook is your call. Zapier, Google Sheet, Supabase, your own database. Doesn't matter. To get AI to write both tags, give it: → The URL parameters you want to capture → Screenshots of your dataLayer from GTM Preview mode on your conversion page → Where you want the data sent Takes about 10 minutes. Paste both tags into GTM as Custom HTML, set your triggers, done. Now every order or lead has first-touch and last-touch attribution tied to actual revenue. Not GA4 session data. Not platform-reported conversions. Real money matched to the campaign that drove it. No dev team. No Shopify requirement. No monthly fee. Works on any site that has GTM installed. Happy to help if you have questions.

@Replit Can't wait to try this

Introducing Replit Animation Vibecode your next viral video in minutes, powered by Gemini 3.1 Pro. (This video was 100% made in Replit Animation)

Most marketing attribution tools assume you're on Shopify. But what about companies that built their own app? Your marketing site is on WordPress. Your app is on a separate domain. Someone clicks an ad → lands on your site → clicks through to your app → makes a purchase. GA4 stitches the session. But your order database has zero idea where that customer came from. The tools that exist (Attributer, Triple Whale, Northbeam) either require Shopify or only work with form submissions into CRMs. If your conversion is a purchase in a custom app? You're on your own. I built a system that solves this using only GTM and first-party cookies. No backend dev needed. Sharing the full breakdown this week. Architecture, code, everything — free. Who else is dealing with this?